From the documentation of SOM, I understand that we use SOM to visualize the underlying pattern in the higher dimensional data. However, since SOM is basically a neural network, so do we need to supply higher dimensional discriminative features extracted from data or the rough data directly and SOM will extract discriminative features automatically. An example of rough data could be the time domain or frequency domain responses of the pristine and defective mechanical structures.
Dear Asif,
Indeed, you raised an interesting query but many times it goes without even saying. I would prefer to use "raw data" definitely after quality testing, since SOM is a unsupervised clustering algorithm.
However, I would suggest you to use both the inputs and report your findings. Typically results should vary much but if it doesn't, try to explore further.
Cheers, A
You can do both depending on the problem you have. but, you should not forget the definition and objective of SOM
SOM is a method of dimensionality reduction from n-dimensions to 2D. Self-organizing maps differ from other artificial neural networks as they apply competitive learning as opposed to error-correction learning such as backpropagation, and they use a neighborhood function to preserve the topological properties of the input space.
SOM reduces the high multidimensionality of data in a lattice without lost the topological properties. Here, it is a case where this poperty is used in the context of optimization:
Article An optimization algorithm guided by a machine learning approach
- Badges
- Science topic
- Similar topics
- Phytochemistry
- Extracts